ANAlYTICAl SCIENCES APRIl 2011, VOl.

27 2011 The Japan Society for Analytical Chemistry

351

Original Papers

Quantitative Analysis of Amino Acids in Dietary Supplements Using Terahertz Time-domain Spectroscopy
Yuko UENO, Katsuhiro AJITO, Naoya KUKUTSU, and Emi TAMECHIKA
NTT Microsystem Integration Laboratories, NTT Corporation, 3-1 Morinosato Wakamiya, Atsugi, Kanagawa 2430198, Japan

We successfully analyzed the concentrations of five amino acids in commercially available dietary amino acid supplements by using terahertz time-domain spectroscopy (THz-TDS) with an error of 12% for the best reproduced components. We also succeeded in analyzing tablets of the supplements wrapped in paper, and thus showed the merit of using THz waves for the nondestructive quantitative analysis of packaged samples by employing the fact that THz waves are capable of passing through several types of packaging material. The ability of THz waves to pass through the paper made it possible to perform a quantitative analysis using the same standard spectra as those used for an unwrapped sample, and the accuracy of a direct quantitative analysis of a packaged sample was almost the same as that of an unwrapped sample. (Received December 17, 2010; Accepted February 23, 2011; Published April 10, 2011)

Introduction
Terahertz time-domain spectroscopy (THz-TDS) is a promising tool in physics and chemistry for characterizing far-infrared vibrational modes, such as rotational, torsional, phonon, inter- and intramolecular modes. Those modes exist in the 0.3 3.0 THz (10 100 cm1) range, which corresponds to the lower end of the far-infrared region.1 One of the attractive properties of THz waves is their ability to pass through a wide variety of materials,2 thus enabling us to see through many types of packaging material, including paper, plastics, leather and wood. This property allows nondestructive inspections of such items as drugs in mail envelopes at post offices.3 The high transparency applies to several molecules that form few interand intramolecular modes, such as hydrogen bonds. Therefore, it becomes possible to achieve the selective detection of certain molecules in mixed samples, which form strong inter- and intramolecular modes in a sample mixed with impurities. Moreover, THz waves cause almost no damage to materials, and so they can be used to detect fragile biological samples49 and to characterize pharmaceutical materials.1013 Thus, THz techniques are attracting increasing interest not only in basic research, but also for practical applications in a variety of fields. Recently, the usefulness of the quantitative analysis of chemicals which have characteristic absorption in the THz region, such as amino acids, by using THz-TDS has been demonstrated;1417 however, the analysis of amino acids in actual foods or drugs has not yet been tested. In this paper, we describe the quantitative analysis of five amino acids in an actual dietary amino acid supplement by using THz-TDS. The supplement contains other ingredients, such as sugars and vitamins, and thus is a suitable sample for testing the selective detection of an amino acid using THz-TDS. We also analyzed tablets of a

commercially available dietary amino acid supplement wrapped in paper, and compared the accuracy of the direct quantitative analysis of the packaged samples with that of unwrapped samples. Since one characteristic of THz waves is their ability pass through commonly used packaging materials, such as paper and plastics, the accuracy with which they can provide a nondestructive quantitative analysis of packaged samples is of interest. The merit of using THz-TDS for the direct quantitative analysis of packaged samples was also shown by comparing it with Raman spectroscopy.

Experimental
Sample preparation Commercial dietary amino acid supplements were purchased from two different makers. The samples of the two different kinds of supplements are denoted as Supplements A and B here. The concentrations of the amino acids and other ingredients were calculated based on nutrition fact sheets and are summarized in Table 1. The supplement granules were ground into powder, and about 80 mg of this powder was compressed into a tablet 10 mm in diameter. The mechanically determined tablet thicknesses were 0.7 and 0.8 mm for Supplements A and B, respectively. In order to eliminate from the spectra the effect of the multiple reflections that occur between the two surfaces of a sample tablet (etalon artifacts), we removed the signal of the multiple reflections prior to Fourier transformation of the time-domain signal to obtain the frequency-domain spectrum. These thickness values provided a sufficient path length for us to eliminate from the spectra the effect of the multiple reflections. Since windowing the data can introduce incorrect information at low frequencies, we do not use the spectra in 0 to 0.5 THz region for further analysis. The final weights of the tablets were also measured to determine the quantity of each amino acid that they contained. Wrapped samples for use in the nondestructive inspection of a packaged tablet were prepared by

To whom correspondence should be addressed. E-mail: ueno.yuko@lab.ntt.co.jp

352
Table 1Concentration of amino acids and other ingredients in Supplements A and B Concentration, % Supplement A Amino acids l-Arg l-Glu l-Ile l-leu l-Val Other ingredients Maltitol Citric acid Sugars Natural flavors Vitamin C Niacin Pantothenic acid Vitamin B6 Vitamin B2 Vitamin B1 Vitamin A Vitamin D Vitamin E 17.3 18.0 12.3 15.3 10.3 Unknown Unknown Unknown Unknown 0.63 0.13 0.08 0.01 0.01 0.01 <0.01 <0.01 0.54 Supplement B 7.5 7.5 10.0 15.0 10.0 Unknown Unknown Unknown Unknown 1.25 0.21 0.06 0.02 0.02 0.01 <0.01 <0.01 0.13

ANAlYTICAl SCIENCES APRIl 2011, VOl. 27

using the 180 backscattering geometry, and the grating in the polychromator was 1200 line/mm. The exposure times for the Raman measurement were 10 to 30 s. The spectral resolution was <2 cm1. We performed all of the measurements at room temperature. Method of quantitative analysis Before obtaining the standard spectra of the major constituents (l-Arg, l-Gln, l-Ile, l-leu, l-Val, maltitol, and citric acid monohydrate), we calculated the concentrations of the constituents in each standard tablet by using the following equations;
C= W w , ( w + wpe )M V

(1) (2)

V = r2d,

where C, W, w, wpe, M, d and r are the concentration, the weight of the tablet, the weight of the mixed amino acid powder, the weight of the mixed PE powder, the molecular weight of the amino acid, and the thickness and radius of the tablet, respectively. The absorption spectra of a sample, A(), were calculated using
A( ) = ln Esam ( ) E ref ( ) .
2

(3)

placing the above tablets of Supplements A and B between sheets of weighing paper (~30 m thick). To obtain the standard THz spectra, that is, the molar absorption coefficient in a certain THz region, standard tablets of the major constituents (l-arginine (l-Arg), l-glutamine (l-Gln), l-isoleucine (l-Ile), l-leucine (l-leu), l-valine (l-Val), maltitol, and citric acid monohydrate) were prepared by mixing a certain amount of each chemical in powder form with high-density polyethylene (PE) powder (Aldrich). Measurement We measured the THz spectra by using a THz-TDS2004 (Aispec) combined with a near-infrared pulse laser to generate optical pulses with a duration time of 10 fs and a wavelength of 800 nm (Integral, FEMTOlASERS). The repetition frequency was 80 MHz. We used a 50-V biased low temperature-grown GaAs (lT-GaAs) photoconductive antenna for both the generation and detection of THz pulses. Other optical set-ups are described in detail elsewhere.14 A sample tablet was attached at the center of the rear of the holder (f7 mm) so as to allow the THz beam to pass through it. The data for the sample were collected by a detector. The reference data consisted of THz pulses that passed through the same aperture in the absence of a tablet. An accumulation number of 32 was used for both the sample and the reference to improve the S/N. The measurement time was within 5 min. The spectral resolution was 0.96 cm1. All of the measurements were acquired at room temperature under a pressure of 50 Pa to avoid water vapor absorption. We measured the Raman spectra using an NIR Raman microprobe system (InVia Reflex, Ranishaw). The excitation light source was the 785-nm line of a laser-diode continuous-wave laser (Ranishaw). The approximately 30 mW NIR laser light was focused on the sample with an objective lens (50, NA 0.55). The laser power at the sample was about 3 mW for all of the measurements. The laser was focused on top of a tablet of amino acid supplement or amino acid powder. The spot size diameter and the focal depth on the sample were about 1 m and <10 m, respectively. The scattered light was collected by

By Fourier transforming the time-domain signal, we obtained the amplitude spectra, Esam() and Eref(), for a sample of thickness d placed in the beam and an empty reference, respectively. We removed the signal of multiple reflections prior to Fourier transforming. Since windowing the data can introduce incorrect information at low frequencies, we did not use the spectra in 0 to 0.5 THz region for further analysis. Our previous study14 showed that the A() resulting from the absorption of the sample is proportional to the concentration, and so A() can be expressed by Beers law, A() = ()Cd, (4)

where () is the molar absorption coefficient. Thus, we calculated the () values of the standard chemicals by averaging the () curves for several different concentrations of standard tablets to obtain a standard spectrum. We used the standard spectra set (l-Arg(), l-Glu(), l-Ile(), l-leu(), l-Val(), maltitol(), and citric acid()) for further quantitative analysis. We can calculate the spectrum (Scalc()) for any mixing ratio of l-Arg, l-Gln, l-Ile, l-leu, l-Val, maltitol, and citric acid monohydrate by using the standard spectra set, Scalc() = kl-Argl-Arg() + kl-Glul-Glu() + + kcitric acidcitric acid(), (5)

where kl-Arg, kl-Glu, kl-Ile, kl-leu, kl-Val, kmaltitol, and kcitric acid are the coefficients for the constituents. To obtain the coefficients that correspond to the most likely ratio of the amino acids in Supplements A and B, we minimized the error between the calculated and experimental spectra.

Results and Discussion

Supplement A, unwrapped The solid line in Fig. 1(a) shows the THz absorption spectrum

ANAlYTICAl SCIENCES APRIl 2011, VOl. 27

353
Table 2Calculated concentration and recovery ratio of amino acid in Supplements A and B Concentration, % Calculated Nutrition sheet Unwrapped Recovery Wrapped Recovery ratio ratio Supplement A l-Arg l-Glu l-Ile l-leu l-Val Supplement B l-Arg l-Glu l-Ile l-leu l-Val 17.3 18.0 12.3 15.3 10.3 7.5 7.5 10.0 15.0 10.0 15.3 19.6 12.2 24.4 13.3 19.5 10.3 9.0 6.6 10.3 0.88 1.09 0.99 1.59 1.29 2.60 1.37 0.90 0.44 1.03 16.7 17.8 16.4 7.7 1.7 19.9 10.5 12.8 38.4 11.7 0.97 0.99 1.33 0.50 0.17 2.65 1.40 1.28 2.56 1.17

Fig. 1THz spectra of unwrapped Supplement A tablet (a) and tablet wrapped in a weighting paper (b). Experimental (bold solid line) and calculated (bold dotted line) spectra are compared. Spectra of the constituents are overlaid to show their contributions to the calculated spectra.

of Supplement A in the 0.5 to 3.0 THz region. Several characteristic peaks were clearly observed. The dotted line shows the calculated spectrum obtained by adding the standard spectra, which were scaled in proportion to the calculated concentrations of five amino acids and maltitol. The spectra of the five amino acid constituents and maltitol are overlaid to show their contributions to the calculated spectra. The experimental spectrum was well reproduced by the calculated spectrum. Figure 3(a) and Table 2 summarize the concentrations of the five amino acids shown in the nutrition fact sheet and calculated results for Supplement A. The recovery ratios (calculated concentration over concentration given in nutrition fact sheet) revealed that the estimation errors, defined by ((recovery ratio) 1) 100, were within 12% for the three constituents (l-Arg, l-Gln, and l-Ile). For l-leu and l-Val, the estimation errors were 59 and 29%, respectively. The order of the concentration was correctly predicted, expect as regards l-leu. Thus, despite employing a rough analysis that did not take account of the contribution of minor components, such as vitamins, we successfully analyzed the actual supplement by using THz-TDS. Supplement B, unwrapped The solid line in Fig. 2(a) shows the THz absorption spectrum of Supplement B in the 0.5 to 3.0 THz region. Several characteristic peaks were clearly observed, and the spectral shape was different from that of Supplement A. The dotted line shows the calculated spectrum obtained by adding the standard spectra, which were scaled in proportion to the calculated concentration of five amino acids, maltitol, and citric acid. The order written in the nutrition fact sheet for Supplement B

Fig. 2THz spectra of unwrapped Supplement B tablet (a) and tablet wrapped in a weighting paper (b). Experimental (bold solid line) and calculated (bold dotted line) spectra are compared. Spectra of the constituents are overlaid to show their contributions to the calculated spectra.

indicates that the percentage of citric acid is between those of and l-Val, although the concentration was not stated. We took citric acid into consideration in the analysis. The spectra of the five amino acid constituents, maltitol, and citric acid are overlaid to show their contributions to the calculated spectra. The experimental spectrum was well reproduced by the calculated spectrum. Figure 3(b) and Table 2 summarize the concentrations of the five amino acids shown in the nutrition fact sheet and the calculated results for Supplement B. The
l-leu

354

ANAlYTICAl SCIENCES APRIl 2011, VOl. 27

Fig. 3Comparison of concentrations of five amino acids provided by a nutrition fact sheet (black), calculated results of unwrapped (gray) and wrapped (hatch) samples of Supplements A and B.

Fig. 4Raman spectra of five amino acids, and unwrapped Supplements A and B.

recovery ratios showed that the estimation errors were within 10% for l-Ile and l-Val. For l-Glu and l-leu, the estimation errors were +37 and 56 %, respectively. The estimation error for l-Arg was as large as +160%. Since the amino acid concentration in Supplement B is smaller than that in Supplement A, and the concentrations of the other ingredients in Supplement B were mostly larger than that in Supplement A, the errors were larger in the quantitative analysis of the amino acids in Supplement B. The results indicate that the detection limit of THz spectroscopy is currently around a few percent, that is, about 0.01 to 0.1 M/mm2 for the molecules whose molecular size is about 102, and the accuracy of the quantitative analysis, is also several percent. This means that, unfortunately, THz spectroscopy is incapable of trace analysis and has no advantage in terms of the detectable range over other spectroscopic methods. However, advances in optical design, brighter light sources and more sensitive detectors may improve the detection limit, and thus make THz spectroscopy a more powerful analytical tool. Supplement A, wrapped We also measured wrapped samples of Supplements A and B by placing a tablet between sheets of weighing paper (~30 m thick) as an example of the nondestructive inspection of a packaged tablet. The solid line in Fig. 1(b) shows the THz spectra of wrapped tablets of Supplement A in the 0.5 to 3.0 THz region. We found that wrapping in paper had very little effect on the spectral shape because the differential spectra calculated by subtracting the absorption intensity of wrapped sample from that of unwrapped sample was satisfactory flat. The peak positions and spectral shapes did not change, although the absorbance increased slightly in the overall frequency (the ratio of the absorption intensity of the difference spectra to that of unwrapped sample was 0.11 at ~1.7 THz), and thus the noise

was larger at frequencies above ~2.5 THz. Thanks to the ability of THz waves to pass through paper, we can perform a quantitative analysis by using the same standard spectra as those used for the unwrapped sample. The dotted line shows the calculated spectrum obtained by adding the standard spectra, which were scaled in proportion to the calculated concentrations of five amino acids and maltitol. We did not use the noisy region above ~2.5 THz for the calculation. The experimental spectrum was well reproduced by the calculated spectrum. The recovery ratios in Table 2 show that the estimation errors were within 33% for the three constituents (l-Arg, l-Gln, and l-Ile). For l-leu and l-Val, the underestimation errors were 50 and 83%, respectively. Since l-Ile and l-Val both have peaks at around 1.7 THz, the estimation of l-Ile and the underestimation of l-Val can occur simultaneously by confusing the two components, which have similar THz absorption profiles. Using a wider spectral region for an unwrapped sample or changing the measurement temperature to see the difference in the spectral shapes will help to avoid confusion and allow us to improve the accuracy. Supplement B, wrapped The solid line in Fig. 2(b) shows the THz spectra of wrapped tablets of Supplement B in the 0.5 to 3.0 THz region. Again, we found that the effect of wrapping in paper on the spectral shape was very small, and the peak positions and the spectral shapes did not change. Due to the small overall frequency absorption, the noise was sufficiently small to allow us to observe a similar region to that for the unwrapped sample, we used the region up to 2.7 THz for the calculation for both the wrapped and unwrapped samples. The dotted line shows the calculated spectrum obtained by adding the standard spectra, which were scaled in proportion to the calculated concentrations of the five amino acids, maltitol, and citric acid. The experimental

ANAlYTICAl SCIENCES APRIl 2011, VOl. 27

Table 3Comparison of the Raman peak positions of Supplements A and B with those of amino acids Supplement A Supplement B
l-Arg l-Gln l-Ile l-leu l-Val

355

418 458 456 489 539 558 577 614 626 655 541 551 557 544 455 480 490 538

404 458

534

542

665 671 711 750 777 827 837 848 875 671 715 776 837 849 779 827 850 882 897 925 947 965 994 1003 1033 1034 1037 1054 1068 1086 1084 1093 1101 1124 1131 1129 1098 1136 1166 1172 1189 1181 1190 1191 1205 1240 1258 1265 1274 1286 1299 1319 1331 1344 1351 1401 1424 1437 1455 1456 1477 1508 1515 1514 1514 1514 1452 1451 1456 1454 1419 1298 1310 1330 1342 1352 1398 1419 1333 1333 1310 1329 1341 1355 1359 1398 1409 1420 1429 1399 1298 1316 1322 1332 1345 1352 1263 1273 1191 1187 1181 1193 1202 1137 1131 1126 1147 1083 1068 1023 1035 905 925 949 965 904 923 928 922 965 983 993 1004 1032 1035 925 947 965 950 966 850 853 874 837 847 675 711 749 670 716 754 776 826 850

778

Fig. 5Raman spectra of wrapped Supplement A and a weighting paper.

the degree of the estimation error was little changed by the wrapping. It is noteworthy that THz-TDS can realize almost the same level of accuracy as quantitative analysis for both packaged and unwrapped samples. lastly, we measured the Raman spectra of Supplements A and B and the five constituent amino acids (Fig. 4). Several sharp peaks were observed. The six large peaks are indicated with dashed lines and were observed in the spectra of both Supplements A and B. Table 3 compares the Raman peak positions. Considering the bandwidth, the peaks observed in the 5 cm1 region were regarded as being the same vibrational mode, and so are placed on the same rows in the table. The six marked peaks in Fig. 4 are shown as gray rows. We found that the contributor to the peaks cannot be identified as a single amino acid component, but is assigned as the sum of the peaks of two or more amino acids. Although a detailed analysis can reveal the degree to which each amino acid contributes to a certain peak, quantitative analysis based on a scattering method, such as Raman spectroscopy, is complicated because calibrations are often non-linear due to the variation in the intrinsic scattering efficiency with the concentration.18 We also measured samples of Supplement A after placing the tablet between sheets of weighing paper. Figure 5 compares the Raman spectra of the wrapped Supplement A (top) and the weighing paper (bottom). The results indicate that the Raman spectra cannot detect any signals from the chemicals in Supplement A through a paper whose thickness (~30 m) exceeds the focal depth (<10 m). Therefore, we confirmed that compared with Raman spectroscopy THz-TDS is a powerful tool for the direct quantitative analysis of packaged samples when the concentration range of the target chemicals exceeds a few percent, that is, about 0.01 to 0.1 M/mm2 for the molecules whose molecular size is about 102.

spectrum was well reproduced by the calculated spectrum. The recovery ratios revealed that the estimation errors were +40, +28, and +17 for l-Glu, l-Ile, and l-Val, respectively. The estimation errors for l-Arg and l-leu were as large as +165 and +156%. Despite the significant change in the result for l-leu,

Conclusions
We successfully analyzed five amino acid concentrations in actual dietary amino acid supplements with an error of 12%

356
for the best-reproduced components. The detection limit with THz spectroscopy was around a few percent, and the accuracy of the quantitative analysis was also several percent. In addition, we succeeded to analyze tablets of commercially available dietary amino acid supplements wrapped in paper. The wrapping had a negligible effect on the spectral shape, and the peak positions and spectral shapes remained unchanged. The ability of THz waves to pass through the paper made it possible to perform a quantitative analysis using the same standard spectra as those used for the unwrapped sample. It is noteworthy that the accuracy of a direct quantitative analysis for a packaged sample was almost the same as that for an unwrapped sample. We confirmed that THz-TDS is a more powerful tool for the direct quantitative analysis of packaged samples than Raman spectroscopy, when the concentration range of the target chemicals exceeds a few percent, that is, about 0.01 to 0.1 M/mm2 for the molecules whose molecular size is about 102.

ANAlYTICAl SCIENCES APRIl 2011, VOl. 27

6. P. F. Taday, I. V. Bradley, and D. D. Arnone, J. Bio. Phys., 2003, 29, 109. 7. K. Yamamoto, K. Tominaga, H. Sasakawa, A. Tamura, H. Murakami, H. Ohtake, and N. Sarukura, Biophys. J., 2005, 89, l22. 8. B. M. Fischer, M. Hoffmann, H. Helm, R. Wilk, F. Rutz, T. Kleine-Ostmann, M. Koch, and P. U. Jepsen, Opt. Express, 2005, 13, 5205. 9. K. Yamamoto, M. H. Kabir, and K. Tominaga, J. Opt. Soc. Am. B, 2005, 22, 2417. 10. P. F. Taday, Phil. Trans. R. Soc. Lond. A, 2004, 362, 351. 11. K. l. Nguyen, T. Fricic, G. M. Day, l. F. Gladden, and W. Jones, Nat. Mater., 2007, 6, 206. 12. H. Wu, E. J. Heilweil, A. S. Hussain, and M. A. Khan, Int. J. Pharm., 2007, 343, 148. 13. J. Aaltonen, K. C. Gordon, C. J. Strachan, and T. Rades, Int. J. Pharm., 2008, 364, 159. 14. Y. Ueno, R. Rungsawang, I. Tomita, and K. Ajito, Anal. Chem., 2006, 78, 5424. 15. Z. Zhang, X. Yu, H. Zhao, T. Xiao, Z. Xi, and H. Xu, Opt. Commun., 2007, 277, 273. 16. R. Nishikiori, M. Yamaguchi, K. Takano, T. Enatsu, M. Tani, U. Chandimal de Silva, N. Kawashita, T. Takagi, S. Morimoto, M. Hangyo, and M. Kawase, Chem. Pharm. Bull., 2008, 56, 305. 17. A. R. Taulbee, J. A. Heuser, W. U. Spendel, and G. E. Pacey, Anal. Chem., 2009, 81, 2664. 18. N. Everall, Analytical Applications of Raman Spectroscopy, ed. M. J. Pelletier, 1999, Chap. 4, Blackwell Science, Oxford, 127.

References
1. Y. Ueno and K. Ajito, Anal. Sci., 2008, 24, 185. 2. D. M. Mittleman, M. Gupta, R. Neelamani, R. G. Baraniuk, J. V. Rudd, and M. Koch, Appl. Phys. B, 1999, 68, 1085. 3. K. Kawase, Y. Ogawa, and Y. Watanabe, Opt. Express, 2003, 11, 2549. 4. E. Pickwell and V. P. Wallace, J. Phys. D: Appl. Phys., 2006, 39, R301. 5. B. M. Fischer, M. Walther, and P. U. Jepsen, Phys. Med. Biol., 2002, 47, 3807.